Creators AISubscribe free
← All insights
Tutorials & How-ToCreators AI Research

How to keep up with AI: Sources & Learning Tips

The best AI newsletters, podcasts, leaders + how to build your learning system in Cursor

CA
Creators AI EditorialDec 9, 2025 · 11 min read
Newsletter artwork for “How to keep up with AI: Sources & Learning Tips”

Hello friends!

We are glad you follow the AI evolution with us. However, we are now at a point where it is next to impossible to keep up with all AI updates and developments. Even I, after working with AI every day and studying it at university, still feel overwhelmed at times.

But here’s the good news: if you ever wondered if it is too late to dig in deeper into AI and make it part of your everyday life, the answer is no. It is never too late.

The key is finding the right resources, making it easier.

Today, we’re delving into

  • What role AI plays in our future
  • AI tools that help to learn AI
  • People and channels worth following
  • How to structure your AI Summarizer in Cursor

Keep your mailbox updated with practical knowledge & key news from the AI industry!

Why It Matters

You’ve seen plenty of times that AI is surrounded by common buzzwords: “fastest-growing”, “innovative”, “revolutionary”, and “game-changing”.

At the same time, each new AI tool brings its own kind of guilt 🫩: the guilt of not exploring enough, of falling behind, of not knowing what others seem fluent in.

People keep putting off projects because they can become obsolete in a couple of months anyway, lowkey stop learning, and just get lost in the noise.

But we’re mentioning it in all our articles, but I’ll say it again — the right AI tools save you time, save you money, and, most importantly, free up your brain space to actually work on creative ideas.

The core basics

If you’re genuinely ready to start an in-depth learning into AI, here’s what that actually looks like:

1. Core AI Concepts Think machine learning algorithms, model building, deep learning basics, and understanding those mysterious “black boxes”.

Check out LLMs Explained Simply & How They Can Save You 800 Hours This Year

2. Specialization Areas You pick your lane: Natural language processing, computer vision, AI for business applications. Then you work on real-world projects. For example, I studied how AI is being used in the political sphere, which honestly opened my eyes to both the possibilities and the risks.

Check out Vibe Designing: AI Workflows with Figma & Others and How AI Can Boost SMB

3. Ethics and Legal FrameworksBecause with great AI power comes great responsibility. Seriously, though, understanding the ethical implications and legal boundaries is crucial. We’re talking bias in algorithms, data privacy, accountability, the whole nine yards.

4. Foundational Technical Skills

Python programming, linear algebra, probability, statistics, and data manipulation. Gonna be honest, this part can be boring, but it’s what separates people who use AI tools from people who actually understand and can build with them.

You might be thinking that this all sounds like way too much work. And you will not be alone, because according to data from LinkedIn, 51% of professionals say that learning AI feels like a second job, and 41% say the pace of change is affecting their well-being.

But today, we’re going to focus more on vibe-learning – getting the general feel and understanding of what’s happening without burning out.

AI Tools to Learn AI

Claude Code and Cursor

Cursor and Claude Code were clearly built for developers, but we shouldn’t underestimate their power for learning. I’ve already dived deep into how they can transform managing finance, marketing, and analytics workflows.

But Cursor can also work as your AI knowledge Base + news analyzer. You can upload documents, course materials, all your notes, references, and make Cursor analyze it all at once. Remember those @links I mentioned? They let you drop context in seconds:

Docs + trends

@docs “HuggingFace blog” → summarize latest model releases → impact ranking

Competitor analysis

@web Mistral vs xAI funding news → timeline + implications

Knowledge debt tracker

@composer build a script that scans my notes → identifies outdated concepts → creates a priority list of topics to review

While in the post about Claude Code, I broke down how insanely efficient the Claude Code + Obsidian combo can be. We can leave behind digging through endless files and worrying about their size, because everything runs locally.

  • Claude Code reads, writes, and automates your workflow without that annoying grunt work.
  • Obsidian keeps everything in plain Markdown files, and this way, you can build out this beautifully connected knowledge base that grows with you.
The feedback on posts about these two brothers from another mother was honestly amazing, so if you haven’t explored those posts yet, you’re missing out on some serious productivity hacks.

Google Studio

Google AI Studio lets you play around with hands-on experiments using models and data. There are many guides on it, but in general, it’s convenient when you want to check out models doing their thing in real-time.

How to use it:

  • Fire up ready-made demos and see how different models handle fresh data
  • Learn to tweak model parameters and watch the output shift before your eyes
  • Run experiments with text, images, or audio without dealing with environment setup
  • Create your own datasets and test how models handle your specific use cases

Google Skills

Have you already heard about Google Skills?

They launched a free AI learning platform, and it’s perfect for step-by-step learning. It’s mainly built of bite-sized lessons and practical exercises covering everything from basic to intermediate AI topics.

How to use it:

  • Pick whatever topic you need right now – ML basics, working with text models, visual models, whatever, and just go through short modules one after another
  • Take quizzes to lock in the material and actually check if you’re getting it
  • Google Skills works great in 10–15 minute daily chunks

NotebookLM

Another Google creation, but I’m separating it because it’s built specifically for self-directed learning. NotebookLM doesn’t make things up, but actually pulls information directly from the resources you upload.

If you’re paranoid about accuracy and want to extract real research gems from your materials, NotebookLM will work for you.

It’s incredible for synthesizing information from multiple sources to understand complex topics and spot trends. For example, you could feed it a bunch of recent model release notes, research summaries, and product updates from different labs. And then when asking any (stupid) question, you can be sure that the model doesn’t hallucinate.

You will find how to customize NotebookLM and learn the basics of building an LLM with it in this post: NotebookLM Updated Guide: Work & Learning Tips

Hugging Face Learn

They have already popped in our digests — open-source AI courses from the creators of Transformers, which are free and hands-on.

How to navigate:

  • Agents Course to build and deploy AI agents
  • Deep RL Course on reinforcement learning with Hugging Face libraries
  • Computer Vision Course to practice with vision models
  • Audio Course to try out transformers for audio

Key Resources to Follow

To get quality information, you need to make sure the people you follow are actually deep into AI.

Influencers on X/LinkedIn

  • Demis Hassabis – CEO DeepMind - Nobel winner, AGI insights​
  • Geoffrey Hinton – Godfather of AI
  • Andrej Karpathy – AI researcher who previously served as the Director of AI at Tesla and was a founding member of OpenAI
  • Sam Altman  –  CEO of OpenAI
  • Ian Goodfellow – practical ML research​
  • Jensen Huang  –  CEO of Nvidia, a pioneer in AI hardware and GPU technology for deep learning
  • Andrew Ng – Founder of DeepLearning.AI
  • Yann LeCun – Chief AI Scientist at Meta
  • Louie Peters –  Co-founder & CEO at Towards AI
  • Fei-Fei Li – AI Researcher & Professor at Stanford University
  • Thomas Wolf  – Co-founder and Chief Science Officer at Hugging Face

Even if you don’t stalk their pages daily, the algorithm will throw their posts into your feed whenever something big drops. That’s how it works for me anyway, and it’s clutch for staying on top of expert takes without actually trying.

Educators and YouTube channels

Podcasts

  • The AI Podcast – AI’s impact on science and technology.
  • Last Week in AI – data science and AI through critical thinking
  • AI Today – AI insight thought leaders, leading technology companies, pundits, and experts
  • How I AI – practical tutorials/workflows
  • AI for Humans – entertaining news/trends
  • Practical AI – complex AI topics and real-world use cases for practitioners
  • AI in Business - practical AI adoption and use cases for business leaders

Reddit Communities

Reddit’s basically ground zero for AI people swapping knowledge. If you look for something practical, these subreddits help you:

Hey, we’re not sleeping on this either and we’re building out our Reddit as well. Come hang out with us there – CLICK!

Newsletters

Learning Assistant in Cursor: Tutorial

You know how we all love saving reels and tweets with recipes or book lists that we promise to read later and never open again. Saving creators’ works the same way. Adding them to a list will not make you better at AI.

Read more about this jack-of-all-trades tool How Cursor Can Be Your AI Assistant & Knowledge Base

The challenge isn’t that of finding interesting AI papers or tools; it’s creating a system.

🫣 The Problem: You have multiple documents and web pages with valuable information, but reading everything manually takes hours. You want a system that can consolidate key insights, bullet points, and actionable items into an organized summary.

🤓 The Solution: A Cursor Summarizer pipeline that:

  • Accepts multiple sources, both local files and web pages.
  • Extracts text and content automatically.
  • Generates individual summaries for each source.
  • Creates a master summary with main takeaways, recurring themes, and contradictions.

Create the Folder Structure

raw - this is where all your input materials are stored

summaries - this is where the Summarizer will save the results

Prompt:

Create this folder structure:
data/
 raw/
outputs/
 summaries/
src/
 pipeline/
 loaders/
Inside src/pipeline/, create a placeholder file called summarizer.py.
Inside src/loaders/, create placeholder files called file_loader.py and web_loader.py.
Make sure the project is ready to accept input files from data/raw/ and save summaries to outputs/summaries/.

In the picture, there is the structure I ended up with, a framework on which I will build my analyses.

Configure Cursor to Work with the Project

In the main Cursor window, type:

@composer I want to build an AI Learning Tracker system.
Create a basic project structure and placeholder files inside src/.
I will be using the folders: data/raw for inputs and outputs/summaries for outputs.

Add Materials to Your Folders

  • If you already have information, notes, saved articles, just drag and drop any PDF / DOCX / TXT into: data/raw/
  • To add a webpage as a source, you have two options to choose from.

+ You don’t download anything and just give the URL to the Summarizer.

    Summarize this page:
    https://example.com/article
    Store the result in outputs/summaries/example_article.md

+ Download HTML / TXT yourself

If you want to keep a local copy of the article:

Right-click → New File → data/raw/article1.html

The Summarizer will process this file.

Run the Summarizer on any FILE

Here’s an example with an article Is Vibe Coding Safe? Benchmarking Vulnerability of Agent-Generated Code in Real-World Tasks

Summarize the file located at:
ata/raw/report.pdf @NAME_OF_THE_FILE

Use a detailed summary mode:

- bullets

- key ideas

- insights

- action items

Save output to:

outputs/summaries/report_summary.md

Check out the picture, Cursor:

✓ loaded the file ✓ ran the pipeline ✓ saved the final markdown file

Run Summarization on a FOLDER (multiple files)

Run a batch summarization.

Take all files from:
@ data/raw/

For each file:
- create individual summary
- save each to outputs/summaries/

Then create one combined master_summary.md with:
- main takeaways across all documents
- recurring themes
- contradictions
- important insights

Cursor will be: ✓ Markdown ✓ Nicely structured ✓ With headings ✓ With bullet points ✓ Ready to copy into Google Docs

So now you’ve got a system where one simple @ call can grab and analyze your files, without losing track of data, and even pull in online content automatically.

And to make life even easier, you can dump summaries into folders like these:

Categories:

  • Key models: LLMs, diffusion, multimodal stuff
  • Specific libraries: vLLM, Triton, LoRA, etc.
  • Agent systems: how companies actually use agents to run workflows, ops, research, and internal stuff
  • Photo & video generation: new tools for marketing, product workflows, synthetic content, and automation
  • Companies & open-source teams: who’s pushing AI forward
  • Standards & specs: OpenAI OPE, Open Model Spec, ONNX, MLPerf
  • Competitors: what your peers and rivals are using
  • Research areas: stuff that could shake up your industry

I set this up once, and now everything is organized, searchable, and ready to use — no chaos or stress.

❗️Do not forget to create a .cursorrules file to set rules for Cursor.

Now that your workspace is ready, you can trigger a full daily research scan with just one command inside Cursor: /daily (or /weekly – change everything as you wish)

# CURSOR RULES - SUMMARIZER WORKSPACE

## Workspace Purpose
This workspace is designed for managing, summarizing, and analyzing documents and web content using Cursor. 

When the user types “/daily”:

1. Collect all sources from:
   - The URLs directly listed in the /daily prompt
   - data/raw/daily/ folder
   - Any URLs found in documents the user has recently opened

2. Use the web loader to fetch readable text from each URL.

3. Summarize each source individually:
   - short summary
   - key takeaways
   - notable quotes
   - links preserved

4. Produce a combined master report.
5. Save the result to:
   outputs/daily_reports/YYYY-MM-DD-daily.md

Formatting rules:
- Markdown only
- No generic filler
- Strong structure with headings and bullet points
- Always place sources at the bottom

Never ask follow-up questions unless critical.

Conclusion

Healthy learning environments don’t force you into “use AI or fall behind” mode. They give you room to be confused, to test things, to step back when your brain is full.

Getting better at AI isn’t about cramming more information, but about picking the stuff that actually fits where you want to go. And that’s the idea we try to support in our newsletters.

AI learning shouldn’t feel like an extra shift after work. It should be something you grow into at your own pace. You stop drowning in updates and start catching insights that matter.

Follow a few people who add real value, skim newsletters that don’t waste your time, and join communities where you get answers instead of hype.

Leave a comment

Archive note

This article was first published in the Creators AI newsletter. View the original edition.

Keep exploring

More in Tutorials & How-To.